Data
I take data nobody hands over and turn it into the decision, the artifact, or the instrument.
Every growth decision I make and every AI system I build runs on data, so when the data is wrong the decision and the system are wrong with it.
Shipped for:OlmiKuCoinArtradeQuraniumEntangleBitazza
[01] WHERE THE DATA COMES FROM
Every market sizing, target list, dashboard and press figure further down this page starts with data somebody had to go and get. Nobody publishes most of those numbers, so I go and get them.
Nine kinds of source, in three groups. I register every source with how much weight it can carry and how old I let it get, and a run stops dead when a core source passes its deadline.
[02] HOW I TURN DATA INTO ANSWERS
Raw data answers nothing on its own. Fifteen operations turn it into the four answers a buyer actually asks for: what a target looks like, which targets to go after, what the market rate is, and which numbers will not hold.
I write each operation in plain code wherever the answer is a count or a rule. A model comes in only where the judgment is the work, like reading a video or folding a scattered footprint into one briefing.
[03] ONE TARGET TORN DOWN
Press RUN. A target's name goes in and one run pulls its whole operation apart: the backend every client shares, how each client is wired, the roster it never announced, the fees in its own terms, and the person who owns the decision.
Nothing here is staged: the tools and the steps are the ones I run. I read only what is public or mine, I grade every field by how sure I am, and the finished dossier opens below. The target is a composite, so it names no real company.
a platform vendor and its embedded clients
- CERTS
- HEADERS
- ROSTER
- CAPTURE
- PACKAGE
- FLAG
- TERMS
- PEOPLE
- DOSSIER
FINDINGS
[04] WHAT COMES OUT
Every job I run ends in a file you can open: a brief that leads with the decisions, a stat sheet a journalist can quote and defend, a board a leadership team reads on a Monday. Each figure inside sits next to the sample it came from and the source behind it, so whoever I hand it to can forward it without rewriting a word.
Open any figure here and it shows its sample, its source, and how I treated it, set small beside the number. Every artifact wears that rule, so the read survives a fact check after it leaves me.
[05] WHERE IT GOES
The store, the transforms and the rules stay the same for every job. Only the decision at the end changes.
Choose a function and I route the same pipeline to it, then show you the call it unblocks, the artifact I hand over, and one measured example of the work.
[06] WHY THE NUMBER HOLDS
I'll show you the one figure in a report I'd flag before I show you the headline. Every number I hand over wears its sample and its source, I grade what I'm sure of and mark what I'm not, and I state what the data can't answer. You can check any of it yourself.
Every figure on this page ships wearing its sample, its source, and how I treated it, set small beside the number so you read the figure and its receipts in one look.
[07] DATA RECORD
Every figure my data work produced sits here once, next to the client it came from.
Each one names the source behind it where I have it, so you can check me.
[08] GET STARTED
Let me turn your data into a decision.
[09] FAQ
The doubts still worth raising after the record, each answered straight, with the place on this page that proves it.
I assemble sources most teams never reach: public filings and registries, platform APIs, my own telemetry, and a target's own artifacts under static analysis. A cost-capped actor is one class among them. Then I write the transforms in plain deterministic code and grade every field. What comes out is a synthesis, and scraping is a small part of the intake.
Anything that is not mine I read under static analysis, with its signature verified first and nothing executed. I use my own account and session at normal-use volume, and public proxies only. I never claim access to anyone's internal dashboards, and I do not tear down a company I am advising.
What I hand over is openable and owned: the store and its transforms, the measurement plan, the dashboards, the briefs, the playbook. The rules that govern them sit in a written standard I sign. Your team reads the code, refreshes the sources on their own schedule, and keeps producing without me in the loop.
I build the store, the transforms, and the deliverables to be operated by whoever owns the data. I work alongside the analysts you already have, wire the sources and the reporting, and hand over files and code they can read. What I install gives them a bigger surface to work from, and it stays theirs to run.
When the evidence is thin, I mark it thin. Every deliverable ends with a section on what the data cannot answer, stated in plain words and never fudged. I have refused to fabricate a figure the data would not support, so what I cannot prove is written down instead of guessed.
A data story built for earned media carries no named comparison. I keep comparative prices and head-to-head claims inside private targeting work, where they belong, and I exclude folklore statistics by rule. What reaches a reporter is built to survive a fact check: real figures, each wearing its sample and its source.